Research On Moving Load Identification Based On BP Neural Networks

Recent years bridge engineering accidents occurred frequently in the world, which resulted in great concern on the bridge structural health. The actual conditions of engineering structures would change when it was used for a period of time, such as material aging and fatigue; as a result, health monitoring is to be a hot problem in bridge engineering that many experts put their emphasis on. As the foundation of health monitoring, moving load identification is an important basis for policy-maker carrying on the valuation and decision of the structure safety. And assessing the physical condition of bridge structure correctly, promise dependable grounds for the bridge structure safety.In this text, the theory and methods of moving load identification was introduced. Based on this, a new method of moving load identification was carried out. Artificial neural networks were used to identify moving load acting on the bridge structures. Setting time to be a dimension of input data of the BP networks, load which changes with the time -called moving load was deserved. Founded on analysis of dynamical displacement under moving loads, BP neural networks for moving load identification were used to identify loads on freely supported beam and continuous beam.Major content of this text as follows:1. Simplified a freely supported beam to a Euler-Bernoulli Beam, a forced vibration equation under moving loads was established. According to the model superposition principle, dynamic response of the beam was obtained. Equation for moving load identification by dynamic response was deduced.2. Analytic solution to dynamic response of freely supported beam under moving load was obtained by means of numerical calculation. Dynamic response of mid-span under different load forms (such as single load, double loads, constant load and time-varying load) was analyzed. Dynamic response of different locations of the beam was also analyzed.3. Time and dynamical response used as input data of the BP networks, moving load used as output data, such BP networks were established to identify moving loads on freely supported beam and continuous beam. Values for the key parameters of the networks such as the step length of time were optimized by numerical imitation for several times, and relationship between the parameters and the identification effect was discussed.Research technique and findings of this text distributed some reference for moving load intelligent identification of bridge structures, and may be the basis for damage identification and health monitoring.